[extropy-chat] Smalley, Drexler and the monster in Lake Michigan

[Robert J. Bradbury]

On Sun, 7 Dec 2003, Robin Hanson wrote:

> So they [goverment officials] want to declare that
> "nanobots" are impossible, with no more specific common definition of what
> the impossible things are than whatever it is that the public is afraid
> of.  As long as respectable scientists can be recruited who say they are
> impossible, this is the position the government agencies will take.  And
> given the vast money available, it was pretty sure that some respectable
> scientists would be found to take this position.  [snip]

I think people are missing a couple of important issues here.  Right now
a Drexler type assembler arm probably is *effectively impossible*.
[And one can comfortably ignore nanorobots.]  The reasons are poorly
developed methods and very high costs.

Lets see:
MECHANOSYNTHESIS: has been done on a limited scale with AFMs and
1-2 atoms.  No parallel AFMs in production to my knowledge.  There
is the Zyvex rotapod but it needs tool tips and needs to be scaled
down from MEMS to nanoscale.  That will happen probably if they
manage to make it through the decade and 5-20 nm litographic/MEMS
methods are developed.  (Right now we are looking at 65 nm circa 2006).

SYNTHETIC CHEMISTRY: it isn't sophisticated enough yet and the computational
capacity isn't up to the job.  The largest synthetic molecules that have
been made are things like Vitamin B-12, vancomycin and maitotoxin.  They
have atomic sizes ranging from ~200-500 atoms.  Even for the fine motion
controller which is 2600 atoms in size but probably only 1/500th to 1/1000th
of an 4,000,000 atom assembler arm design (not counting the motors and
computer interface you might need to control it).  [The reason the ratios
aren't exact is there is a fair amount of repitition in the assembler
arm design.]  Then you only have 2 people in the world (Merkle and Drexler)
who have ever designed nanoscale parts -- where does the rest of the
design come from?  Finally to assemble something with 4,000,000+ atoms
you are going to require *lots* of chemical reactions -- potentially
millions.  Even if you had the design in hand it is probably far too
complicated for the human mind and the current retrosynthetic analysis
programs are only capable of handling molecules from a few hundred
to a few thousand molecules in size.  The problem is the exponential
growth in the number of reactions you have to try as the nanoparts
increase in size.

ENZYMATIC CHEMISTRY: The problem here is that even though we are starting
to accumulate databases with thousands of enzymes (due to genome sequencing)
we don't know many of the structures yet -- though our rate of determining
them should increase significantly over the next couple of years.  Our
computer predictions of protein structure from amino acid sequence are
still pretty poor and our ability to design proteins has only begun to
develop in a few groups over the last 5-8 years.  And very few people
are working on de novo enzyme design.

Finally there is the issue of COST.  I've looked at this for ENZYMATIC
CHEMISTRY in [1].  You can disagree with my analysis but I would bet
they would not be low by more than an order of magnitude and its even
less likely that they are high.  The cost to design *just* the
enzymes for the Fine Motion Controller to do the assembly of smaller
chemical building blocks -- *if* we had lots of designers and some
reasonable skill in that process would be $5.8 million dollars.
For a nanoassembler arm the cost is $9 billioin dollars and for a
single nanorobot its $17 trillion dollars.  These figures do not
include the costs of the design of the nanoparts (nanosubcomponents)
of the finished nanomachines.  Since these amounts are significantly
beyond what most government government grants are for, at this time
you would have to recruit 5-10 groups to tackle the problems if
you really wanted to deal with them seriously.  That is the real
problem with Smalley's position -- an insufficient number of good
scientists have read Drexler's literature enough to want to take
a position against him.  That would allow us to attempt to enroll
the types of team leaders that can both bring home the money.

That's why one of the projects I'd like to work on is a complete
retrosynthetic analysis of the fine motion controller.  Its the
smallest nanopart Merkle & Drexler designed.  Its within 5x
of what has been done to date.  Once that has been accomplished
Smalleys position immediately starts to fall apart and the race
is on.

Now the good news is that natural trends in molecular biology,
synthetic chemistry, biochemical knowledge, computation, etc.
seem to be driving the costs down, in part due to the increase
in the knowledge base, better methods, working at smaller scales,
etc.  By 2010 things might start to look feasible for "moon shot"
type approaches and by 2020 things may really start to get feasible.

So I'm not so sure that there is a lot of politics involved (though
government officials might the impression out to the public that
they don't want to develop things that are dangerous -- but you
know that isn't true for DARPA and the pentagon.  The problem is
that most of the NNI work is being done by the NSF, NIH and
NASA which are probably a little more risk averse.

Robert